In Vivo Autopolysialylation and Localization of the Polysialyltransferases PST and STX*

A select group of mammalian proteins have been shown to possess a 2,8-polysialylated oligosaccharide chains. The best studied of these proteins is the neural cell adhesion molecule (NCAM). Polysialylation of NCAM has been shown to decrease NCAM-dependent and independent cell adhesion. PST (ST8Sia IV) and STX (ST8Sia II) are the two polysialyltransferases responsible for NCAM polysialylation. Recent studies revealed that PST itself is autopolysialylated in vitro (Muhlen-hoff, M., Eckhardt, M., Bethe, A., Frosch, M., and Ger-ardy-Schahn, R. (1996) EMBO J. 15, 6943–6950). Here we report studies on the biosynthesis and localization of the PST and STX polysialyltransferases. Both PST and STX are expressed as high molecular mass, polydisperse forms that are associated with the cell and found soluble in the medium. Analysis of these high molecular mass forms by glycosidase digestion and serial immunopre-cipitation/immunoblot experiments demonstrated that PST and STX are autopolysialylated in vivo . Indirect immunofluorescence microscopy and immunoprecipitation analyses demonstrated that autopolysialylated PST and STX are localized in the Golgi, on the cell surface, and in the extracellular space. The cell surface and extracellular localization of these polysialylated polysialyltransferases suggest that their polysialic acid chains, like those of NCAM, may modulate cell interactions. Polysialic acid is

Polysialic acid is a linear homopolymer of N-acetyl or Nglycolyl neuraminic acid (Neu5Ac or Neu5Gc) occurring in ␣2, 5-, 8-, or 9-ketosidic linkages (reviewed in Ref. 1). Polysialic acid has been found widely distributed throughout nature, from the capsular polysaccharide of neuroinvasive bacteria, such as Escherichia coli K1 (2), to unfertilized rainbow trout eggs and the jelly coat of sea urchin eggs (3,4), to human tumors and tissues (1,5). While there has been a great deal of research on bacterial polysialylation, eukaryotic, especially mammalian, protein polysialylation has been the focus of intense investigation in recent years.
Although polysialic acid has been found throughout the taxonomic spectrum, only the oligosaccharides of a few mammalian proteins bear this modification. While polysialic acid has been found on the ␣-subunit of the rat brain voltage-sensitive sodium channel (6) and unidentified proteins in breast cancer and basophilic leukemia cell lines (7), the most abundant carrier of polysialic acid is neural cell adhesion molecule (NCAM) 1 (8). Polysialylated NCAM has been observed in developing brain (9), kidney (10), heart, and muscle (11). It is widely postulated that the polysialylation of NCAM oligosaccharides during the development of the nervous system and other organ systems in the embryo and neonate leads to a general decrease in cell adhesion (12)(13)(14)(15)(16)(17). It is believed that the presence of cell surface polysialic acid disrupts the homophilic binding properties of NCAM and facilitates cellular migration, neurite outgrowth, and synaptic plasticity (12)(13)(14)(15)(16)(17). Interestingly, polysialylated NCAM is also reexpressed on some metastatic cancers such as neuroblastoma (18), small cell lung carcinoma (19), and the highly metastatic kidney tumor, Wilms tumor (5). As in development, cell surface-expressed polysialylated NCAM is thought to increase the migration of cancer cells, thereby enhancing their metastatic potential (18, 20 -23).
Two recently cloned polysialyltransferases have been shown to be responsible for the polysialylation of NCAM. The first, STX (ST8Sia II), is a type II membrane protein with a predicted molecular mass of 42.5 kDa and six potential N-linked glycosylation sites. STX has been cloned from rat (24), mouse (25), and human (26) sources and was shown to have polysialyltransferase activity toward N-linked oligosaccharide structures (25)(26)(27). The second polysialyltransferase, PST (ST8Sia IV), is also a type II membrane protein with a predicted molecular mass of 41.2 kDa. PST has five potential N-linked glycosylation sites, with four reportedly being used (28). PST was cloned from human (29), hamster (30), mouse (31), rat (32), and chicken 2 sources and was also shown to have polysialyltransferase activity toward N-linked oligosaccharides (30,31,34). Interestingly, recent in vitro studies on PST revealed that the PST enzyme is itself modified by polysialic acid and that this modification may be important for activity (28).
In light of the observation that PST is polysialylated in vitro, no published data to date have shown in vitro polysialylation of STX or in vivo polysialylation of either enzyme. Here we report the in vivo autopolysialylation of both PST and STX. Pulsechase immunoprecipitation analyses revealed that both enzymes are associated with the cell and are found as high molecular mass, soluble forms in the extracellular space. Glycosidase digestion demonstrated that the high molecular mass forms of PST and STX are due to modification of the enzymes' complex, N-linked oligosaccharides. This modifica-tion was subsequently identified as ␣2,8-polysialic acid by its susceptibility to endo-N-acetylneuraminidase (Endo N) digestion (35) and the immunoprecipitated proteins' reactivity with the anti-polysialic acid antibody, OL.28 (7), which has been shown to recognize ␣2,8-polysialic acid chains of more than 4 units 3 . Last, indirect immunofluorescence localization of autopolysialylated PST and STX shows that both localize to the Golgi and the cell surface. Based on these data, we hypothesize that autopolysialylated PST and STX, found on the cell surface and in the extracellular space, may modulate the interactions of expressing cells.

Methods
Construction of Epitope-tagged PST and STX-Full-length PST cDNA was obtained from Dr. Minoru Fukuda (Burnham Institute, La Jolla, CA) and full-length STX cDNA was obtained from Dr. John Lowe (University of Michigan, Ann Arbor, MI). The PST and STX cDNAs were amplified by polymerase chain reaction amplification using Vent DNA polymerase and oligonucleotide primers specific for each cDNA (GATATCCAAGATGCGCTCCATTAGGAAG and TCTAGACCTT-GCTTTACACACTTTCCTG for PST; GATATCATGCAGCTGCAGTTC-CGGAG and TCTAGACCCGTGGCCCCATCGCACTG for STX). These primers specifically introduced an EcoRV restriction site at the 5Ј-end and an XbaI site at the 3Ј-end of each cDNA. Following restriction enzyme digestion, PST and STX cDNAs were ligated into previously digested pcDNA3.1/V5-His vector DNA to complete the construction of PST-V5 and STX-V5. Inserts were confirmed using the Sequenase version 2.0 DNA Sequencing Kit (Amersham Pharmacia Biotech) and the T7 primer TAATACGACTCACTATAG (Genosys Biotechnologies, The Woodlands, TX).
Transfection of COS-1, CHO, and Lec2 CHO Cells-COS-1, CHO, and Lec2 CHO cells maintained in DMEM, 10% FBS (COS-1) or ␣-minimal essential medium, 10% FBS (CHO and Lec2 CHO) were plated on 100-mm tissue culture plates or 12-mm glass coverslips and grown in a 37°C, 5% CO 2 incubator until 50 -70% confluent. Lipofectin CHO cells with V5 epitope-tagged PST or STX cDNA and expression of these enzymes in the cells for 18 h, 100-mm tissue culture dishes of transfected cells were incubated with cysteine/methionine-free DMEM for 1 h. After incubation, this medium was replaced with 3.5 ml of fresh cysteine/methionine-free DMEM containing 100 Ci/ml 35 S-Express protein labeling mix (NEN Life Science Products). Cells were incubated with the radiolabel for 1 h at 37°C in a 5% CO 2 incubator. After labeling, medium was removed, cells were washed, and the labeled proteins were chased for various periods of time with DMEM, 10% FBS. Cell medium was collected, and cells were washed with 10 ml of PBS and lysed in 1 ml of immunoprecipitation buffer 2 (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40, 0.1% SDS).
Both PST and STX enzymes were immunoprecipitated from cell lysates and media at each time point using 2.5 g of anti-V5 epitope tag antibody and protein A-Sepharose (Amersham Pharmacia Biotech), as described previously (36). However, to avoid breakdown of the polysialic acid, the boiling step was omitted, and the immunoprecipitation beads were resuspended in 50 l of Laemmli sample buffer containing 5% ␤-mercaptoethanol and directly loaded into the gel wells. Immunoprecipitated proteins were separated on 7.5% separating, 3% stacking SDS-polyacrylamide gels (37). Radiolabeled proteins were visualized by fluorography using 10% 2,5-diphenyloxalzole in dimethyl sulfoxide (38), and gels were exposed to Kodak BioMax MR film at Ϫ80°C.
Glycosidase Digestions of Immunoprecipitated Enzymes-COS-1 cells expressing either PST-V5 or STX-V5 were metabolically labeled for 1 h and chased for 6 h, and radiolabeled proteins were immunoprecipitated as described above. Following the final wash, the immune complexes bound to protein A-Sepharose beads were treated with glycosidases for 18 h with shaking at 37°C. For PNGase F, 1500 units of enzyme was added to 70 l of distilled H 2 O and 30 l of 1ϫ reaction buffer (50 mM sodium phosphate, pH 7.5) per manufacturer's instructions. PNGase F cleaves between the innermost N-acetylglucosamine and asparagine residues of high mannose, hybrid, and complex oligosaccharides from N-linked glycoproteins. For V. cholerae Nm digestions, 30 milliunits of enzyme was added to 70 l of 1ϫ reaction buffer (50 mM sodium acetate, pH 5.5, 4 mM calcium chloride, 100 g/ml bovine serum albumin) per manufacturer's instructions. V. cholerae Nm hydrolyzes terminal N-or O-acylsialic acids that are linked via ␣2,3-, ␣2,6-, or ␣2,8-bonds. For Endo N digestions, a 1:10 dilution of enzyme in 20 mM Tris-HCl, pH 7.4, 0.2 mg/ml bovine serum albumin was added to the immune complexes. Endo N cleaves ␣2,8-polysialic acid chains at random sites, requiring a minimum chain length of five sialic acid residues for activity (35). For Endo H digestions, 10 milliunits of enzyme was added to 90 l of 1ϫ reaction buffer (0.1 M sodium citrate, pH 6.0, 0.075% SDS, 0.2% ␤-mercaptoethanol) (39). Endo H cleaves only high mannose or hybrid N-linked oligosaccharide structures of glycoproteins. Digested samples were electrophoresed on a 7.5% SDS-polyacrylamide gel after the addition of 50 l of Laemmli buffer (37) and ␤-mercaptoethanol to 5% final concentration. Radiolabeled proteins were visualized by fluorography, and gels were exposed to Kodak BioMax MR film at Ϫ80°C.
Serial Immunoprecipitation and Immunoblotting of Polysialyltransferase and Treatment of Immunoblots with Endo N-COS-1 cells were transfected with PST-V5 or STX-V5 cDNA, and unlabeled cell lysates and medium were collected after an overnight incubation. Lysates and medium were subjected to serial immunoprecipitation by rotation of samples for 1.5 h with 0.75 g of V5 antibody and 50 l of a 50% slurry of protein A-Sepharose in PBS. After this incubation, the immune complexes were centrifuged, and supernatants were removed to new tubes. An additional 0.75 g of V5 antibody and 50 l of a 50% slurry of protein A-Sepharose in PBS were added to the supernatants, and this mixture was rotated for an additional 1.5 h. This scheme was repeated a total of four times, with the final supernatant being retained. The immune complexes were electrophoresed on 7.5% separating, 3% stacking SDS-polyacrylaminde gels. Following electrophoresis, proteins were electrophoretically transferred to nitrocellulose membranes overnight at 500 mA. The membranes were processed for immunoblotting according to the manufacturer's protocol (Pierce). Anti-V5 epitope tag antibody (IgG) was diluted 1:5000, and anti-polysialic acid antibody, OL.28 (IgM), was diluted 1:500 in blocking buffer (5% dry milk in Tris-buffered saline, pH 8.0, 0.1% Tween 20 for anti-V5 epitope tag antibody; 2% dry milk in Tris-buffered saline, pH 8.0, for OL.28 antibody). Horseradish peroxidase-conjugated secondary antibodies, goat anti-mouse IgG and goat anti-mouse IgM, were each diluted 1:8000 in blocking buffer (5% dry milk in Tris-buffered saline, pH 8.0, 0.1% Tween 20). Immunoblots were developed using the SuperSignal chemiluminescence kit (Pierce) and exposed to Kodak BioMax MR film at room temperature.
Separately, aliquots (2.5% of total volume) of the final supernatants 3 K. Kitajima, unpublished results. of cell lysates and medium from the serial immunoprecipitation were electrophoresed on a 7.5% SDS-polyacrylamide gel, and total protein was transferred to nitrocellulose membranes. The immunoblots were cut into strips, and selected strips were treated overnight at 37°C with a 1:10 dilution of Endo N. Immunoblotting with the anti-V5 epitope tag and anti-polysialic acid antibodies was performed as described above.
Immunofluorescence Localization of PST and STX-COS-1 cells were plated on glass coverslips, transfected with PST-V5 or STX-V5, and processed for immunofluorescence microscopy as described previously (40). Briefly, cells were treated with either Ϫ20°C methanol to visualize internal staining or 3% paraformaldehyde to visualize cell surface staining. Anti-V5 epitope tag antibody was diluted 1:100, and the OL.28 anti-polysialic acid antibody, fluorescein isothiocyanate-conjugated secondary antibodies, goat anti-mouse IgG and goat anti-mouse IgM, were diluted 1:200 in 5% normal goat serum/PBS blocking buffer prior to use. Coverslips were mounted on glass slides using 20 l of mounting medium (15% (w/v) Vinol 205 polyvinyl alcohol, 33% (v/v) glycerol, 0.1% azide in PBS, pH 8.5). Cells were visualized and photographed using a Nikon Axiophot microscope equipped with epifluorescence illumination and a 60ϫ oil immersion Plan Apochromat objective.

PST and STX Are Modified to High Molecular Mass Sialylated Forms That Are Found Associated with the Cell and
Soluble in the Extracellular Space-In addition to their expected localization in the Golgi, many glycosyltransferases are found as soluble forms in the extracellular space and at low levels on the cell surface (41)(42)(43)(44)(45)(46). To determine whether PST and STX are cleaved and secreted into the extracellular space, we performed pulse-chase immunoprecipitation analyses. PST and STX tagged with the V5 epitope at their carboxyl termini were transiently expressed in CHO cells. Cells were metabolically labeled with 35 S-Express protein labeling mix for 1 h and chased with unlabeled medium for 0 -12 h. V5 epitope-tagged proteins were immunoprecipitated from both cell lysates and media with the anti-V5 epitope tag antibody, and immunoprecipitated proteins were analyzed on SDS-polyacrylamide gels. An approximately 57-kDa protein was immunoprecipitated from lysates of CHO cells expressing the PST-V5 enzyme (Fig.  1A). The size of this protein corresponds well with the expected molecular mass of the full-length epitope-tagged PST (PST-V5) modified by four N-linked oligosaccharides (28). A smaller 52-kDa protein was immunoprecipitated from media and is likely to represent a cleaved and secreted form of the enzyme. More interestingly, we observed a heterogeneous, immunoreactive band extending from 105 to 190 kDa associated with the cell at early chase time points (1 and 3 h) and in the medium from 1 to 12 h of chase (Fig. 1A). The polydisperse appearance of this high molecular mass immunoreactive band suggests that a large proportion of the PST-V5 is highly glycosylated and most likely polysialylated.
Analysis of STX-V5 protein transiently expressed in CHO cells revealed a similar pattern of biosynthesis. A protein of 58 kDa molecular mass was observed associated with the cells, and a smaller, 53-kDa species appeared in the medium after 1 h of chase (Fig. 1C). Again, the size of the 58-kDa protein corresponds well with the calculated molecular mass of the epitope-tagged STX (STX-V5) modified by four N-linked oligosaccharides, while the size of the 53-kDa protein suggests that it is a cleaved and secreted form of the full-length enzyme. Like PST-V5, we also observed a heterogeneous band extending from 70 to 131 kDa in the medium from 1 to 12 h of chase (Fig.  1C). The polydisperse appearance of this band again suggests that the high molecular mass form of the STX-V5 is highly glycosylated and most likely polysialylated. These results demonstrate that both PST and STX are cleaved and secreted from cells as both highly modified and less modified forms. The polydisperse appearance of the highly modified forms of PST-V5 and STX-V5 suggest that both enzymes are extensively glycosylated and probably polysialylated.
In order to determine what proportion of the molecular mass of PST-V5 and STX-V5 could be attributed to the addition of sialic acid, we expressed the PST-V5 and STX-V5 proteins in Lec2 CHO cells, metabolically labeled the cellular proteins, and immunoprecipitated the enzymes with the anti-V5 epitope tag antibody (Fig. 1, B and D). Lec2 CHO cells lack a functional CMP-NeuAc transporter and consequently show a very significant reduction in sialylation of glycoproteins and glycolipids compared with wild type CHO cells (47,48). The PST-V5 enzyme secreted from the Lec2 cells migrated on SDS-polyacrylamide gels with a molecular mass ranging from 55 to 82 kDa (Fig. 1B). This is markedly smaller than the 105-190-kDa PST-V5 secreted from wild type CHO cells (Fig. 1A). Similarly, the STX-V5 enzyme secreted from Lec2 cells migrated on SDSpolyacrylamide gels with a molecular mass of 56 -80 kDa, and this too was significantly smaller than the 70 -131-kDa STX-V5 protein secreted from wild type CHO cells (Fig. 1, compare C  and D). These data indicate that the expression of the high molecular mass forms of PST-V5 and STX-V5 polysialyltransferases depend upon the cells' ability to sialylate glycoproteins.
PST and STX Are Polysialylated on Complex N-Linked Oligosaccharides-The previous results suggest that the bulk of these enzymes' molecular mass increase could be attributed to the addition of sialic acid residues to the oligosaccharide structures of PST-V5 and STX-V5. In light of Gerardy-Schahn and colleagues' (28) demonstration that PST is autopolysialylated in vitro, it is likely that the high molecular mass forms of the polysialyltransferases we observe are a result of their in vivo autopolysialylation. To determine whether PST-V5 and STX-V5 are autopolysialylated on complex N-linked oligosaccharides, we analyzed their carbohydrate structures using PN-Gase F, Endo H, Endo N, and V. cholerae Nm. Radiolabeled PST-V5 and STX-V5 were immunoprecipitated with the anti-V5 epitope tag antibody from CHO cell medium after 6 h of chase. The radiolabeled protein-antibody complexes were bound to protein A-Sepharose beads and then treated overnight with specific glycosidases. Treatment of soluble PST-V5 with PNGase F reduced the high molecular mass and low molecular mass soluble forms of the enzyme to two species of 40 and 36 kDa. This demonstrated that the high molecular mass form of PST-V5 possesses highly modified N-linked oligosaccharides and that the low molecular mass form of the enzyme possesses less modified N-linked oligosaccharides (Fig. 2, PST-V5, ϩPNGase F). The predicted molecular mass of unglycosylated PST-V5 is approximately 45 kDa, suggesting that the two species observed after PNGase F digestion of PST-V5 represent different proteolytically processed forms of PST-V5. Treatment of PST-V5 with Endo H had no effect on the high molecular mass form of the enzyme (Fig. 2, PST-V5, ϩEndo H). However, the low molecular mass, soluble form of PST-V5 was susceptible to Endo H digestion, indicating that this form possesses only high mannose N-linked oligosaccharides. Digestion of PST-V5 with either Endo N or V. cholerae Nm reduced the molecular mass of the enzyme to 57-101 kDa (Fig. 2, PST-V5, ϩV. cholerae Nm, ϩEndo N). The sensitivity of the high molecular mass form of the PST-V5 to Endo N, an endoglycosidase specific for ␣2,8-linked polysialic acid (35), demonstrates that this polysialyltransferase is indeed polysialylated itself.
Using an identical approach, analysis of the modification of STX-V5 with these glycosidases gave similar results as those obtained for glycosidase-treated PST-V5. Briefly, PNGase F treatment of STX-V5 resulted in a single band of 37 kDa (Fig.  2, STX-V5, ϩPNGase F), indicating that the high molecular mass modification of STX-V5 resides on the enzyme's N-linked oligosaccharides. Also, like PST-V5, the high molecular mass form of STX-V5 was not sensitive to treatment with Endo H, while the low molecular mass form of the enzyme was sensitive to this endoglycosidase (Fig. 2, STX-V5, ϩEndo H). Digestion of STX-V5 with either Endo N or V. cholerae Nm reduced the apparent molecular mass of STX-V5 to 55-103 kDa (Fig. 2, STX-V5, ϩV. cholerae Nm, ϩEndo N). Taken together, these results demonstrate that the high molecular mass modification of STX-V5 and PST-V5 is the result of sialylation, specifically polysialylation, of the complex N-linked oligosaccharides of both enzymes.
The OL.28 antibody as described under "Methods." As shown in Fig.  3, both PST-V5 and STX-V5 proteins immunoprecipitated from COS-1 cell lysates and media were immunoreactive with the anti-polysialic acid antibody, OL.28 (Fig. 3A, PST-V5 and STX-V5, . These data and the sensitivity of these proteins' oligosaccharides to Endo N (Fig. 2, PST-V5 and STX-V5, ϩEndo N) conclusively demonstrate that both PST-V5 and STX-V5 are autopolysialylated in vivo.
To test the possibility that there are other glycoproteins expressed in COS-1 cells that act as substrates for the polysialyltransferases, we analyzed aliquots of the final supernatant retained from the serial immunoprecipitations with the OL.28 and the anti-V5 epitope tag antibodies (Fig. 3B shows cell lysates only). Immunoblot analysis of the final supernatants from the medium immunoprecipitations demonstrated that the serial anti-V5 antibody immunoprecipitations had completely depleted the medium of both PST-V5-and STX-V5-expressing cells of any polysialylated, OL.28-reactive material (data not shown). Immunoblot analysis of the final supernatants from the cell lysate immunoprecipitations demonstrated that there was residual OL.28-reactive polysialylated material remaining in the final supernatant following the serial immunoprecipitations (Fig. 3B, PST-V5 and STX-V5, OL.28 antibody, ϪEndo  N). Immunoblots of this residual polysialylated material were treated with Endo N to remove polysialic acid and then incubated with the anti-V5 antibody to determine whether this material represented unique polysialylated proteins or inefficiently immunoprecipitated V5-tagged PST or STX (Fig. 3B,  PST-V5 and STX-V5, V5 antibody, ϩEndo N). The anti-V5 epitope tag antibody did recognize protein on the Endo Ntreated PST-V5 and STX-V5 immunoblots, suggesting that polysialylation is masking the V5 epitope in solution (see serial immunoprecipitations; Fig. 3) and on the immunoblots (Fig.   3B, PST-V5 and STX-V5, V5 antibody, ϪEndo N). Based on these results, it is likely that most and perhaps all of this residual polysialylated protein is inefficiently immunoprecipitated autopolysialylated polysialyltransferases.
Autopolysialylated PST and STX Are Localized in the Golgi and on the Cell Surface of COS-1 Cells-Immunoblot analyses of cells expressing either PST-V5 or STX-V5 demonstrated that the bulk of these proteins and polysialic acid are associated with the cells in the steady state (data not shown). To determine where PST-V5 and STX-V5 are located in the cell, PST-V5 and STX-V5 were transiently expressed in COS-1 cells and localized by indirect immunofluorescence microscopy using both the anti-V5 epitope tag and OL.28 anti-polysialic acid antibodies. The anti-V5 epitope tag antibody detected PST-V5 and STX-V5 predominantly in the Golgi apparatus in permeabilized cells (Fig. 4, A and B). In some cells, the enzymes were also observed in the endoplasmic reticulum. The presence of enzyme in the ER has been observed with other glycosyltransferases (45,49,50) and is probably the result of enzyme overexpression. The anti-V5 epitope tag antibody also showed immunostaining of the surface of COS-1 cells expressing PST-V5 or STX-V5 (Fig. 4, C and D). The OL.28 antibody detected polysialic acid in the Golgi, in vesicular compartments, and on the cell surface of permeabilized COS-1 cells expressing PST-V5 or STX-V5 (Fig. 4, E and F). Strikingly, very strong OL.28 cell surface staining was observed for unpermeabilized, PST-V5-and STX-V5-expressing COS-1 cells (Fig. 4, G and H). The differences in intensity of anti-V5 antibody and OL.28 antibody cell surface staining could be due to the presence of multiple polysialic acid chains on a single V5-tagged polysialyltransferase molecule and/or the partial masking of the V5 epitope, as was suggested by the results in Fig. 3. Our results demonstrate that the autopolysialylated polysialyltransferases PST and STX are localized in the Golgi and at the cell surface of transiently expressing cells. The presence of polysialylated PST and STX on the cell surface and in the extracellular space suggest that these polysialylated proteins may modulate cell interactions in vivo.

DISCUSSION
Previous work by Gerardy-Schahn and colleagues demonstrated that PST is autopolysialylated in vitro (28). We wondered whether both PST and STX, the two known polysialyltransferases responsible for NCAM polysialylation, are autopolysialylated in vivo and where they are localized. Immunoprecipitation analyses demonstrated that both PST and STX are found as high molecular mass forms associated with the cell and in the cell medium (Fig. 1). Glycosidase digestions demonstrated that PST and STX are ␣2,8-polysialylated on complex N-linked oligosaccharides (Fig. 2). The immunoreactivity of the immunoprecipitated PST and STX with the anti-polysialic acid antibody, OL.28, again confirmed that these polysialyltransferases are autopolysialylated (Figs. 2 and 3). In addition, serial immunoprecipitation analyses suggest that these enzymes represent the majority of, or even the only, polysialylated proteins expressed by COS-1 cells transfected with PST-V5 or STX-V5 cDNAs (Fig. 3). Finally, indirect immunofluorescence microscopy (Fig. 4) and immunoprecipitation time courses (Fig. 1) demonstrated that PST and STX are localized not only in the Golgi but also at the cell surface and in the extracellular space.
The Endo N sensitivity and OL.28 anti-polysialic acid antibody reactivity of the PST and STX proteins expressed in CHO and COS-1 cells demonstrated that these two polysialyltransferases are polysialylated themselves. Muhlenhoff et al. (28) demonstrated that PST-1 could autocatalytically polysialylate its own N-linked oligosaccharide chains in vitro and that ter- minal ␣2,3or ␣2,6-linked sialic acid was not required for autopolysialylation. In addition, this group showed that asialo-PST-1 synthesized in Lec2 cells, but not agalacto-PST-1 synthesized in Lec8 cells, maintained the ability to polysialylate NCAM in vitro. From these data, they suggest that autopolysialylation of PST-1 may be required for enzyme activity. It is not clear how they came to this conclusion. It is clear that the PST-1 expressed in wild type CHO cells synthesized more polysialylated NCAM in their in vitro assay than did the enzyme synthesized in CHO 6B2 (Lec2) cells that lacked sialic acid on its N-linked oligosaccharides. This may indicate that polysialylated PST-1 has an enhanced activity relative to the unsialylated enzyme. However, the experiments of Muhlenhoff et al. (28) suggested that this was not because the preassembled polysialic acid chains were transferred from the polysialyltransferase to the NCAM acceptor.
In our analysis of the biosynthesis and secretion of PST and STX expressed in wild type CHO and Lec2 CHO cells, we noticed differences in the extent of PST-V5 and STX-V5 polysialylation in wild type CHO cells. When expressed in CHO cells, PST-V5 appeared to be polysialylated to a greater extent than the STX-V5 protein. Similar results were observed when these proteins were expressed in COS-1 cells (data not shown). These observations and those of others (51) suggest that the two polysialyltransferases may have different limits on the length of the ␣2,8-polysialic acid chains that serve as their substrate. We also observed that the enzymes expressed in the Lec2 CHO cells, which lack the ability to sialylate glycopro-teins, were secreted at a lower rate than the enzymes synthesized in wild type CHO cells (Fig. 1, compare A to B and C to D). Likewise, the lower molecular mass population of PST-V5 and STX-V5 proteins synthesized in wild type CHO cells that lacked polysialic acid chains remained cell-associated for longer times than did the high molecular mass polysialylated forms of the polysialyltransferases (Fig. 1, A and C). While sialylation/polysialylation was not a prerequisite for cleavage and secretion, it seemed that sialylation/polysialylation did enhance the rate of enzyme cleavage and secretion. Other researchers have noticed that N-linked oligosaccharide structures do influence protein trafficking in the cell. The addition of N-linked oligosaccharides to a growth hormone-VSV G chimeric protein allowed it to be transported out of the Golgi to the cell surface (52,53), while the presence of oligosaccharide structures on a soluble protein can influence the polarity of its secretion (54). How the polysialylation of PST and STX is influencing the Golgi retention of these enzymes is not clear. However, these observations suggest that PST and STX polysialylation may shorten enzyme residence time in the Golgi and allow them to move more rapidly to the cell surface or another post-Golgi compartment, where they are cleaved and secreted.
The serial immunoprecipitation analyses performed in Fig. 3 show that there is a population of polysialylated enzymes that is not efficiently immunoprecipitated by the anti-V5 epitope tag antibody. After several immunoprecipitations, it appeared that we had depleted the OL.28-reactive, polysialic acid-containing material from the cell lysates of PST-V5-or STX-V5- expressing   FIG. 3. PST-V5 and STX-V5 are autopolysialylated in vivo. A, serial immunoprecipitations using the anti-V5 epitope tag antibody were performed on unlabeled cell lysate or medium fractions from COS-1 cells transiently expressing either PST-V5 or STX-V5 proteins. The immunoprecipitated samples were separated on 7.5% SDS-polyacrylamide gels and subjected to immunoblot analysis using the OL.28 anti-polysialic acid antibody and horseradish peroxidase-conjugated secondary antibodies. B, aliquots (2.5%) of the final supernatants remaining from the serial immunoprecipitations of PST-V5 and STX-V5 from COS-1 cell lysates were separated on 7.5% SDS-polyacrylamide gels and subjected to immunoblot analysis using the anti-polysialic acid antibody OL.28 or the anti-V5 epitope tag antibody. The anti-V5 epitope antibody did not recognize any proteins on the immunoblots until these blots were treated with Endo N to remove ␣2,8-polysialic acid (PST-V5 and STX-V5, V5 antibody, ϩ or Ϫ Endo N). cells (Fig. 3A, PST-V5 and STX-V5, IP#1-4). However, the final supernatant continued to be reactive with the anti-polysialic acid antibody and with the anti-V5 epitope tag antibody only after treatment of the immunoblots with Endo N (Fig. 3B, PST-V5 and STX-V5). These results suggest that much of the residual polysialylated material in the final supernatant of the serial cell lysate immunoprecipitations represents inefficiently immunoprecipitated V5-tagged enzymes. It may be that the extent of polysialylation and/or the specific placement of polysialic acid chains lead to the masking of the V5 epitope. In addition, while we cannot completely rule out the presence of other polysialylated proteins, it seems likely that the polysialyltransferases represent the bulk of polysialylated material expressed by the transfected COS-1 cells, since the major polysialylation substrates, NCAM and the voltage-sensitive sodium channel, are not expressed in these cells (26). 4 We have localized PST and STX to the Golgi and cell surface of COS-1 cells using the anti-V5 epitope tag antibody and the OL.28 anti-polysialic acid antibody (Fig. 4, C and D (anti-V5 antibody) and G and H (OL.28 antibody)). In addition, it is clear that these enzymes, like other glycosyltransferases, are cleaved and secreted from both CHO (Fig. 1) and COS-1 cells (data not shown). Other glycosyltransferases, and in particular sialyltransferases, have been localized in more than one cellular compartment. The ␣2,6-sialyltransferase (ST6Gal I) demonstrates differential localization in different cell types. For example, in the intestinal absorptive cells, this enzyme is found in the trans-most cisternae of the Golgi and at the cell surface (55). In the adjacent intestinal goblet cells, this enzyme is found throughout most of the Golgi cisternae, soluble in mucin droplets, and at the cell surface (55). Recently, we have identified two isoforms of the ST6Gal I that differ in their localization and processing (45). The ST cys isoform is retained intracellularly in the Golgi, while the ST tyr isoform is found in the Golgi and at low levels on the cell surface and is cleaved and secreted into the cell medium (45). The cleavage and secretion of the ST6Gal I and the polysialyltransferases is not surprising and is probably not an artifact of overexpression in a tissue culture cell system, since many soluble glycosyltransferases have been detected in body fluids such as serum, colostrum, and urine (for examples, see Refs. 56 -60).
It is unclear what function cell surface and soluble glycosyltransferases could be performing. It seems unlikely that these cell surface and soluble enzymes act as glycosyltransferases, since their sugar nucleotide donors are not present in the extracellular space. One possibility, previously suggested by others (56,57,61), is that they are utilizing their ability to bind carbohydrates and are acting as lectins that could mediate cell adhesion if correctly positioned at the cell surface. The sialoadhesins (I-type lectins) are a family of sialic acid-specific lectins that are also members of the immunoglobulin superfamily. These cell surface lectins have been demonstrated to mediate the interactions of a variety of cell types via binding of specifically linked cell surface sialic acid residues (reviewed in Refs. 33 and 62). It is possible that cell surface or soluble sialyltransferases could also mediate interactions via binding galactose or sialic acid residues. The presence of long polysialic acid chains on polysialyltransferases could negatively modulate a putative lectin-mediated cell adhesion process, as they do in NCAMmediated cell adhesion. Alternatively, the polysialic acid chains may act as substrates to enhance a sialic acid-specific, lectinmediated, cell adhesion process. The role of these polysialylated polysialyltransferases in mediating or modulating cell adhesion or even cell signaling processes in vivo relies on their cell surface and extracellular expression in normal cells and cancer cells. These and other investigations of how the expression of these polysialylated proteins influence cell adhesion are currently under way.